Last night the Kobo CEO had a longer thread on Bluesky talking about how, when, and where they are using AI and what type of AI they are using.

Short version is that they are against using Generative AI that is trained on authors content without their expressed and explicit consent. But have used small models that help with proper metadata, flag explicit material for proper review, look for malicious duplication or auto-generated AI modifications of existing books or scraped websites.

They also reposted it to their writinglife blog.